PANNZER: high-throughput functional annotation of uncharacterized proteins in an error-prone environment

Area of Science:

• Bioinformatics
• Computational Biology
• Proteomics

Background:

• Protein databases have grown significantly, leading to an increase in unannotated sequences.
• A substantial percentage of protein sequences (32%) are labeled as 'Unknown protein' in databases like UniProtKB.
• Existing functional annotations contain a high error rate (30-40%).

Purpose of the Study:

• To introduce a high-throughput tool, Protein ANNotation with Z-score (PANNZER), for improved protein functional annotation.
• To enhance the reliability and accuracy of predicting protein function.
• To predict both Gene Ontology (GO) classes and free-text functional descriptions.

Main Methods:

• Development of the PANNZER tool for high-throughput annotation.
• Utilizing weighted k-nearest neighbour algorithms.
• Incorporating statistical testing to ensure annotation reliability.

Main Results:

• PANNZER demonstrated superior performance in free-text description prediction compared to all competing methods.
• Significant improvements were observed in Gene Ontology (GO) class prediction, surpassing previous benchmarks like the CAFA 2011 challenge.

Conclusions:

• PANNZER offers a reliable and high-throughput solution for protein functional annotation.
• The tool effectively predicts GO classes and functional descriptions with high accuracy.
• PANNZER represents a significant advancement in bioinformatics tools for understanding protein function.